Stop substituting entities with an empty value

[naitoh]

String#gsub("") matches at every position, so an entity declared with an empty value was inserted between every character when a text node or an attribute value was normalized.
That also kept the other entities from matching the text:

"abc aaa" -> "abc &a;"  # was "∅a∅b∅c∅ ..."

Substituting an empty entity is pointless -- a reference to it and its replacement text are equivalent -- so skipping it loses nothing.

[Copilot]

🟢 Approval recommended

The change is narrowly scoped, corrects a well-defined failure mode, and is covered by targeted regression tests for both text and attribute serialization.

Pull request overview

This PR fixes a normalization bug in REXML::Text.normalize where entities with an empty replacement value caused String#gsub("") to match at every position, resulting in entity references being inserted between every character and preventing other entities from matching.

Changes:

• Skip doctype entities whose replacement value is nil or empty during normalization to avoid gsub("") behavior.
• Add regression coverage for text-node serialization with an empty entity declared in the doctype.
• Add regression coverage for attribute serialization with an empty entity declared in the doctype.

File summaries

File	Description
lib/rexml/text.rb	Skips entities with empty replacement values during normalization to prevent gsub("") inserting references everywhere.
test/test_text.rb	Adds a regression test ensuring text normalization still replaces non-empty entities when an empty entity exists.
test/test_attribute.rb	Adds a regression test ensuring attribute serialization still replaces non-empty entities when an empty entity exists.

Review details

• Files reviewed: 3/3 changed files
• Comments generated: 0
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.